CN110082136A - Rotary machinery fault diagnosis method based on Retrieval method Support Vector Machines Optimized - Google Patents
Rotary machinery fault diagnosis method based on Retrieval method Support Vector Machines Optimized Download PDFInfo
- Publication number
- CN110082136A CN110082136A CN201910318220.1A CN201910318220A CN110082136A CN 110082136 A CN110082136 A CN 110082136A CN 201910318220 A CN201910318220 A CN 201910318220A CN 110082136 A CN110082136 A CN 110082136A
- Authority
- CN
- China
- Prior art keywords
- training sample
- malfunction
- working signal
- support vector
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M99/00—Subject matter not provided for in other groups of this subclass
- G01M99/005—Testing of complete machines, e.g. washing-machines or mobile phones
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a kind of rotary machinery fault diagnosis methods based on Retrieval method Support Vector Machines Optimized, working signal is extracted from different sensors first under each malfunction in rotating machinery, time-frequency characteristics are carried out respectively extracts simultaneously reduction, obtain the feature vector of each working signal, the training sample of each malfunction is obtained based on these feature vectors, it is divided into training sample set A and training sample set B, first using training sample set A as training sample, the processing of parameter pre-optimized is carried out using kernel function and penalty factor of the Retrieval method to more disaggregated models based on support vector machines network, then suboptimization again is carried out to obtain more disaggregated models using training sample set B again, feature vector is extracted from each sensor when rotating machinery breaks down, it inputs more disaggregated models and obtains diagnostic result.The present invention can effectively improve the accuracy and efficiency of rotary machinery fault diagnosis.
Description
Technical field
The invention belongs to Construction Machinery System fault diagnosis technology fields, more specifically, are related to a kind of based on cloud something lost
The rotary machinery fault diagnosis method of propagation algorithm Support Vector Machines Optimized.
Background technique
With the modernization of industry and the rapid development of science and technology, rotating machinery is used as in industry using most extensively
One of equipment, be increasingly being applied to the fields such as electric power, petrochemical industry, aviation and a variety of defense industries.Whirler
Tool equipment constantly develops towards directions such as high speed, systematization and automations, and the scale of production system is gradually increased, machinery knot
Structure also becomes increasingly complex, and interrelated, close-coupled, working performance index are higher and higher between each other for each equipment.It is revolving
Turn in mechanical equipment work operation, along with many uncertain factors, some equipment inevitably generate some failures, and one
The critical component of a certain equipment of denier, which generates failure, can then occur a series of chain reaction, serious also to will cause whole production line
Halt production, in turn result in huge economic loss even casualties.The safety of rotating machinery, it is maintainable and reliable
Property has become the hot spot studied at present, and the peace of rotating machinery can not only be ensured by establishing and improve fault diagnosis technology field
Row for the national games, at the same to improve economic well-being of workers and staff, reduce maintenance cost and ensure personnel safety have positive meaning.
Traditional fault diagnosis technology has less been applicable in complicated rotating machinery structural diagnosis, and to operator's
It is more demanding.With the development of artificial intelligence, intelligent Fault Diagnosis Technique is developed rapidly, and mainly has three categories method: (1)
Fault diagnosis, (2) based on model fault diagnosis and (3) Knowledge based engineering fault diagnosis based on signal processing.Wherein base
Need the assessment of design parameter and working condition with accurate system's model in the fault diagnosis of model, for complicated whirler
Tool equipment, the diagnostic method are uneconomic；Fault diagnosis based on signal processing needs to analyze system acquisition signal
Processing, relative to the fault diagnosis based on model more economically and with certain reliability, but under certain specific environments,
There are some uncertain informations, and certain influence can be generated to fault message, and thus feature extraction problem and breakdown judge need
Further solve；The fault knowledge of history since Knowledge based engineering fault diagnosis, can not exclusively have as human expert
Higher reliability.Meanwhile under current big data era, in practical implementation, existing intelligent Fault Diagnosis Technique
There are speed it is to be improved, accuracy rate is lower and cannot achieve high-precision classification the problems such as, need to research and solve.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide one kind to optimize supporting vector based on Retrieval method
The rotating machinery working signal collected is carried out time frequency analysis processing, passes through cloud by the rotary machinery fault diagnosis method of machine
More disaggregated models based on support vector machines of genetic algorithm optimization realize the diagnosis to rotating machinery fault, can effectively improve
The accuracy and efficiency of rotary machinery fault diagnosis.
For achieving the above object, the present invention is based on the rotating machinery faults of Retrieval method Support Vector Machines Optimized to examine
Disconnected method the following steps are included:
S1: it is intercepted at random from the working signal of different sensors under the R kind malfunction of rotating machinery respectively several
Length is the working signal L of Ms,n(m), wherein s=1,2 ..., S, S indicate number of sensors, n=1,2 ..., Ns, NsIt indicates to come
From the working signal quantity of s-th of sensor, m=0,1 ..., M-1, and M > T, T indicate the period of working signal；Remember every section
Working signal Ls,n(m) corresponding label is Ys,n, label Ys,nFailure shape for the corresponding rotating machinery of identification work signal
State, Ys,n=1 ..., R；
It is special to extract its temporal signatures, frequency domain that time frequency analysis processing is carried out for every section of working signal of every kind of malfunction
Seek peace time and frequency domain characteristics, the quantity of every kind of feature determine according to actual needs, using these features as initial characteristics, then for
The initial characteristics of the working signal of different sensors carry out reduction, the characteristic that s-th of sensor initial characteristics reduction of note obtains
Amount is Ds, it is finally total to obtain K feature, whereinRemember each working signal Ls,n(m) corresponding feature vector
Xs,n=(Xs,n(1),Xs,n(2),…,Xs,n(Ds)), wherein Xs,n(d) working signal L is indicateds,n(m) value of d-th of feature, d in
=1,2 ..., Ds；The working signal feature vector of each sensor is classified according to malfunction label, is obtained s-th
Feature vector set φ of the working signal of sensor under r kind malfunctions,r, r=1,2 ..., R.
S2: for each malfunction, the training sample of the malfunction is obtained in accordance with the following methods respectively: for r
Kind malfunction, from corresponding S feature vector set φs,rIt is middle randomly select respectively a feature vector, X 's,r, then according to
Sensor serial number combines to obtain an assemblage characteristic vector Z=(X '1,r,X′2,r,…,X′S,r)=(z1,z2,…,zK), zkIt indicates
K-th of element of assemblage characteristic vector Z, k=1,2 ..., K, the corresponding malfunction label of assemblage characteristic vector Z are r, i.e. structure
At a training sample；Above procedure is repeated, several training samples are obtained respectively to every kind of malfunction；
All training samples are divided into two parts, a part is used as training sample set A, and another part is as training sample set
B, the training sample that each training sample is concentrated includes all malfunctions；
S3: more disaggregated models of the building based on support vector machines network, input are assemblage characteristic vector, are exported as failure
State tag；
S4: using training sample set A as training sample, using Retrieval method to more points based on support vector machines network
The kernel function and penalty factor of class model carry out the processing of parameter pre-optimized；
S5: the kernel function and penalty factor that step S4 parameter pre-optimized is obtained are as based on the more of support vector machines network
The kernel function of disaggregated model and the initial value of penalty factor, using training sample set B as training sample, using Retrieval method pair
The kernel function and penalty factor of more disaggregated models based on support vector machines network carry out parameter optimization, obtain final more
Disaggregated model；
S6: when rotating machinery breaks down, S sensor is used to collect S length as the working signal of MTherefrom extract the K feature that reduction in step S1 obtainsForm assemblage characteristic vectorIt is input in the trained more disaggregated models of step S5, obtains diagnostic result.
The present invention is based on the rotary machinery fault diagnosis methods of Retrieval method Support Vector Machines Optimized, are rotating first
Working signal is extracted from different sensors under mechanical each malfunction, time-frequency characteristics is carried out respectively and extracts simultaneously reduction, obtain
The feature vector of each working signal obtains the training sample of each malfunction based on these feature vectors, is divided into trained sample
This collection A and training sample set B, first using training sample set A as training sample, using Retrieval method to based on support vector machines
The kernel function and penalty factor of more disaggregated models of network carry out the processing of parameter pre-optimized, then again using training sample set B into
Suboptimization extracts feature vector from each sensor when rotating machinery breaks down to obtain more disaggregated models to row again,
It inputs more disaggregated models and obtains diagnostic result.
The invention has the following advantages:
1) present invention has carried out comprehensive time-frequency characteristics extraction to original working signal, avoids the feature of signal not perfect
Property, while redundancy feature, invalid feature are screened, to obtain and the definitely associated characteristic information of malfunction；
2) present invention is improved for genetic algorithm, using Retrieval method successively to based on support vector machines network
The parameters of more disaggregated models carries out pre-optimized and optimization processing, effectively reduces extra search so that Evolution of Population towards
Optimal direction carries out, that is, reduces the time of diagnosis operation, and improves the accuracy of network diagnosis；
3) it is excellent to have that structure is simple, parameter amount is few for more disaggregated models proposed by the present invention based on support vector machines network
Point, the requirement to hardware resource is low, and economic requirement is low, has certain generalization ability, and the experiment proved that the present invention can be with
Effectively improve the fault diagnosis accuracy rate of rotating machinery.
Detailed description of the invention
Fig. 1 is the specific embodiment stream the present invention is based on the method for diagnosing faults of Retrieval method Support Vector Machines Optimized
Cheng Tu；
Fig. 2 is excellent to more disaggregated models progress parameter based on support vector machines network using Retrieval method in the present invention
The flow chart of change；
Fig. 3 is that the present invention and control methods are accurate for the fault diagnosis of U.S.'s Case Western Reserve University data in the present embodiment
Rate statistical chart；
Fig. 4 is that the present invention and control methods are time-consuming for the fault diagnosis of U.S.'s Case Western Reserve University data in the present embodiment
Comparison diagram；
Fig. 5 is that the present invention and control methods unite for the fault diagnosis accuracy rate of autonomic nerve way platform data in the present embodiment
Meter figure；
Fig. 6 is that the present invention and control methods compare the fault diagnosis time-consuming of autonomic nerve way platform data in the present embodiment
Figure.
Specific embodiment
A specific embodiment of the invention is described with reference to the accompanying drawing, preferably so as to those skilled in the art
Understand the present invention.Requiring particular attention is that in the following description, when known function and the detailed description of design perhaps
When can desalinate main contents of the invention, these descriptions will be ignored herein.
Fig. 1 is the specific reality the present invention is based on the rotary machinery fault diagnosis method of Retrieval method Support Vector Machines Optimized
Apply mode flow chart.As shown in Figure 1, the present invention is based on the rotary machinery fault diagnosis sides of Retrieval method Support Vector Machines Optimized
The specific steps of method include:
S101: working signal data processing:
Several length are intercepted at random from the working signal of different sensors under the R kind malfunction of rotating machinery respectively
For the working signal L of Ms,n(m), wherein s=1,2 ..., S, S indicate number of sensors, n=1,2 ..., Ns, NsIt indicates to come from s
The working signal quantity of a sensor, m=0,1 ..., M-1, and M > T, T indicate the period of working signal；Remember every section of work
Signal Ls,n(m) corresponding label is Ys,n, label Ys,nFor the malfunction of the corresponding rotating machinery of identification work signal, Ys,n
=1 ..., R.
It is special to extract its temporal signatures, frequency domain that time frequency analysis processing is carried out for every section of working signal of every kind of malfunction
Seek peace time and frequency domain characteristics, the quantity of every kind of feature determine according to actual needs, using these features as initial characteristics, then for
The initial characteristics of the working signal of different sensors carry out reduction, the characteristic that s-th of sensor initial characteristics reduction of note obtains
Amount is Ds, it is finally total to obtain K feature, whereinRemember each working signal Ls,n(m) corresponding feature vector
Xs,n=(Xs,n(1),Xs,n(2),…,Xs,n(Ds)), wherein Xs,n(d) working signal L is indicateds,n(m) value of d-th of feature, d in
=1,2 ..., Ds.The working signal feature vector of each sensor is classified according to malfunction label, is obtained s-th
Feature vector set φ of the working signal of sensor under r kind malfunctions,r, r=1,2 ..., R.
S102: training sample is obtained:
For each malfunction, the training sample of the malfunction is obtained in accordance with the following methods respectively: for the event of r kind
Barrier state, from corresponding S feature vector set φs,rIt is middle randomly select respectively a feature vector, X 's,r, then according to sensing
Device serial number combines to obtain an assemblage characteristic vector Z=(X '1,r,X′2,r,…,X′S,r)=(z1,z2,…,zK), zkIndicate combination
K-th of element of feature vector Z, k=1,2 ..., K, the corresponding malfunction label of assemblage characteristic vector Z are r, that is, constitute one
A training sample.Above procedure is repeated, several training samples are obtained respectively to every kind of malfunction.
All training samples are divided into two parts, a part is used as training sample set A, and another part is as training sample set
B, the training sample that each training sample is concentrated includes all malfunctions.The present invention constructs two training sample sets and is easy for
Subsequent parameter Optimization Work, when dividing training sample set can with random division, but due in the present invention disaggregated model use
Based on the disaggregated model of support vector machines network, and in support vector machines network, Margin Vector is for support vector machines network
Training it is extremely important, therefore the vector of near border is placed in training sample set A as far as possible.It is excellent in order to improve subsequent parameter
The efficiency of change proposes a kind of training sample set division methods in the present embodiment:
For each malfunction, calculate the average vector of its all training sample, then calculate each training sample with
Training sample is ranked up by the distance of the average vector from big to small according to distance, and training is added in Q training sample before selection
Sample set A, wherein determine according to actual needs, training sample set B is added in remaining training sample to the value of Q.
S103: more disaggregated models are constructed:
More disaggregated models based on support vector machines network are constructed, input is assemblage characteristic vector, is exported as failure shape
State label.
The fault type of rotating machinery has R kind in the present invention, and characteristic parameter relevant to rotating machinery fault state has K
A, more disaggregated models of the building based on support vector machines network carry out Classification and Identification to the various faults feature of rotating machinery.This
RBF kernel function is used in embodiment, selects the more disaggregated models of one-to-one method construct, and wherein input layer number is K, exports
Node layer number is R.
S104: more disaggregated model parameter pre-optimized:
Using training sample set A as training sample, using Retrieval method to more classification based on support vector machines network
The kernel function and penalty factor of model carry out the processing of parameter pre-optimized.
Retrieval method has the advantages that genetic algorithm, while being improved genetic algorithm, and whole evolution speed is made
Faster, convergence rate is improved, unnecessary search is reduced.Fig. 2 is using Retrieval method in the present invention to based on support
More disaggregated models of vector machine network carry out the flow chart of parameter optimization.As shown in Fig. 2, using Retrieval method pair in the present invention
The specific steps that more disaggregated models based on support vector machines network carry out parameter optimization include:
S201: real coding:
Kernel function and penalty factor to more disaggregated models based on support vector machines network carry out real coding.Traditional
Genetic algorithm is binary coding.With the increase of problem complexity and the raising of precision, the code length of simple generic algorithm
Also it will increase, occupy a large amount of memory, greatly reduce the performance of algorithm.The present invention is improved by the way of real coding
Algorithm performance.
S202: initialization population:
Using the real coding of kernel function and penalty factor as the individual in Retrieval method population, initialization population, kind
Group's size is denoted as Q.
S203: individual adaptation degree is calculated:
For each individual in population, using its corresponding kernel function and penalty factor to based on support vector machines network
More disaggregated models be configured, cross validations, the nicety of grading that will be obtained are carried out to the more disaggregated models using training sample
As individual adaptation degree, it is clear that nicety of grading is bigger, and individual is more excellent.
S204: judge whether otherwise meet termination iterated conditional enters step if it is satisfied, then entering step S205
S206.For the fault diagnosis of rotating machinery, terminate iterated conditional can there are two types of, one is whether to reach maximum evolutionary generation；
Second is that whether fault diagnosis accuracy rate within a certain period of time without significant change, selects as needed in practical applications.
S205: parameter optimization result is obtained:
Optimum individual is selected from current population, using its corresponding kernel function and penalty factor as parameter optimization result.
S206: selection operation:
Individual in population is selected.Selection operation is completed using tournament selection method, it is ensured that individual adaptation degree
It is worth biggish individual and remains progress next step genetic manipulation, to accelerates the evolution of entire population.
S207: crossover operation:
Traditional genetic algorithm is dyed by the individual random pair in population, and with the crossover probability switching part of individual
Body.However, this method makes the Evolutionary direction of genetic algorithm unknown and is unable to control.Cloud model is a kind of uncertain mould
Type, it can be converted between qualitativing concept and quantitative values based on traditional Probability Statistics Theory and fuzzy theory.Its benefit
The numerical characteristics of qualitativing concept and its uncertainty of concept are indicated with three cloud expectation, entropy and super entropy numerical characteristics.Y condition
Cloud generator is cloud generator under the conditions of given three above numerical characteristic and given degree of certainty.It is unnecessary in order to reduce
Search, accelerates evolutionary rate, and crossover operation replaces crossover probability using Y condition cloud model.
Crossover operation in the present invention method particularly includes: be first calculated by linear function and to be selected in step S206
The degree of certainty μ of each parent individuality；Then the numerical characteristic of cloud model is calculated, including cloud it is expected Ex, entropy EnWith super entropy He；Root again
Y condition cloud generator is executed according to three numerical characteristics of the degree of certainty and cloud model that are calculated and obtains a pair of of offspring, calculates institute
The individual adaptation degree of each parent individuality selected in obtained offspring and step S206 selects preferably Q individual.
The calculation formula of degree of certainty μ is as follows:
Wherein, FmaxAnd FminThe overall situation " maximum " and " minimum " fitness value of contemporary population are respectively represented, F ' is intersection two
The greater of father's individual adaptation degree, μmaxAnd μminIt is the maximum value and minimum value of artificially specified degree of certainty.
Cloud it is expected Ex, entropy EnWith super entropy HeCalculation formula it is as follows:
Wherein, ExIndicate cloud expectation, FfAnd FmRespectively indicate fitness, the X of two individual of parentfAnd XmIt respectively indicates and is handed over
Pitch two parent individualities of operation, EnIndicate entropy, xfmaxIndicate the corresponding individual of current population highest fitness, xfminExpression is worked as
The corresponding individual of the preceding minimum fitness of population, HeIndicate super entropy, c1And c2It is artificially specified constant parameter.
E it is expected from cloudxExpression formula can be seen that the linear function of the selection individual of parent two to express desired Ex, this is advantageous
In next-generation closer to the higher side of individual fitness value.Entropy EnDirectly proportional to region of search, entropy is bigger, the coverage area of cloud
It is bigger, so that search range of the individual in crossover operation is bigger.In Y condition cloud generator, Normal Cloud is that one kind is general just
State distribution, generally selects 6≤c in conjunction with the accuracy and speed of evolution algorithm according to the principle of " 3 σ " of normal distribution1≤3p(p
For Population Size).Meanwhile with super entropy HeIncrease, stable tendency can reduce to a certain extent, but if super entropy HeReduce,
Randomness can then be lost.In summary, it should value rationally be carried out to parameter according to the actual situation.
S208: mutation operation:
Mutation operation, return step S203 are carried out to the Q individual that step S207 is obtained.
S105: more disaggregated model parameter suboptimization again:
The kernel function and penalty factor that step S3 parameter pre-optimized is obtained are as more points based on support vector machines network
The kernel function of class model and the initial value of penalty factor, using training sample set B as training sample, using Retrieval method to base
Parameter optimization is carried out in the kernel function and penalty factor of more disaggregated models of support vector machines network, obtains final more points
Class model.
S106: fault diagnosis:
When rotating machinery breaks down, S sensor is used to collect S length as the working signal of M
Therefrom extract the K feature that reduction in step S1 obtainsForm assemblage characteristic vectorIts is defeated
Enter into the trained more disaggregated models of step S105, obtains diagnostic result.
Embodiment
Technical solution and technical effect in order to better illustrate the present invention, using a specific example to work of the invention
Make process and technical effect carries out analytic explanation.Bearing fault is one of rotating machinery typical fault, therefore the present embodiment
The open data of bearing fault that U.S.'s Case Western Reserve University is respectively adopted and the bearing event for independently building the acquisition of fault simulation platform
Barrier data carry out experiment test.
For U.S.'s Case Western Reserve University bearing fault data, choosing 6 kinds of fault types is respectively that (1) rolling element slightly damages
Hurt (B1)；(2) rolling element seriously damages (B2)；(3) inner ring slight damage (I1)；(4) inner ring seriously damages (I2)；(5) outer ring
Slight damage (O1)；(6) (O2) is seriously damaged in outer ring, every kind of fault type is chosen respectively from driving end acceleration sensing
The acceleration vibration signal that device, fan end acceleration transducer and base end acceleration transducer collect.By random
The mode of selection chooses totally 100 groups of samples to every kind of fault type, and every group of data include 1024 sample points.Table 1 is this implementation
Bearing fault information table based on the building of U.S.'s Case Western Reserve University bearing fault data in example.
Table 1
5 kinds of fault types: (1) inner ring slight damage (I1) are respectively set for independently building fault simulation platform；(2) in
It encloses moderate lesion (I2)；(3) inner ring seriously damages (I3)；(4) outer ring moderate lesion (O2)；(5) outer ring slight damage (O1).It is right
The two both vertically and horizontally acceleration transducer HD- for being located at bearing block is randomly selected in every kind of fault type
YD-221 collects acceleration vibration signal (A0, A1) and model WT0180, for measuring the vertically and horizontally side of axis
To each 200 groups of 3 displacement sensors (V0-V2) of information of voltage, every group includes 1024 sample points.Table 2 is the present embodiment
In based on independently build fault simulation platform acquisition building bearing fault database.
Table 2
Time frequency analysis processing is carried out for every section of working signal in the present embodiment.The time domain of working signal is special in the present embodiment
Sign be respectively as follows: root-mean-square value, root amplitude, absolutely square amplitude, the kurtosis factor, shape factor, kurtosis, degree of skewness, pulse because
Son, peak value, nargin coefficient, frequency domain character are respectively as follows: square frequency, gravity frequency, then frequency variance uses wavelet packet analysis
Method carries out time-frequency characteristic analysis to working signal, and E1~E8 totally 8 Parameters of Time-frequency Field are calculated, as time and frequency domain characteristics,
Wherein E1~E8 is the energy of signal to be reconstructed and normalized a kind of later time and frequency domain characteristics parameter after wavelet analysis.
It is total to obtain to 11 temporal signatures available to each working signal, 3 frequency domain characters and 8 time and frequency domain characteristics
22 features.Features above is the feature being commonly used in time frequency analysis, and details are not described herein for specific formula for calculation.
Since the present invention is directed having multisensor, i.e. the rotating machinery of multi-source working signal is needed to each biography
The working signal of sensor extracts feature respectively, therefore total has S × 22 feature.For the data of U.S.'s Case Western Reserve University,
There are 3 sensors, one is obtained 3*22=66 feature, and feature reduction is then carried out to it, uses rough set in the present embodiment
Theory carries out feature reduction.Table 3 is characteristic statistics of U.S.'s Case Western Reserve University data after rough set theory feature reduction
Table.
| Time-frequency characteristics | Time-frequency characteristics | Time-frequency characteristics | |||
| 1 | Drive end _ maximum value | 11 | Drive end _ E4 | 21 | Fan end _ E6 |
| 2 | Drive end _ root-mean-square value | 12 | Drive end _ E5 | 22 | Base end _ root-mean-square value |
| 3 | Drive end _ root amplitude | 13 | Drive end _ E6 | 23 | Base end _ root amplitude |
| 4 | Drive end _ shape factor | 14 | Fan end _ root-mean-square value | 24 | Base end _ kurtosis |
| 5 | Drive end _ pulse factor | 15 | Fan end _ kurtosis | 25 | Base end _ degree of skewness |
| 6 | Drive end _ peak value | 16 | Fan end _ pulse the factor | 26 | Base end _ gravity frequency |
| 7 | Drive end _ gravity frequency | 17 | Fan end _ nargin the factor | 27 | Base end _ E1 |
| 8 | Drive end _ frequency variance | 18 | Fan end _ degree of skewness | 28 | Base end _ E2 |
| 9 | Drive end _ E2 | 19 | Fan end _ E1 | 29 | Base end _ E3 |
| 10 | Drive end _ E3 | 20 | Fan end _ E3 | 30 | Base end _ E4 |
| 31 | Base end _ E7 | 32 | Base end _ E8 |
Table 3
For independently building the data of failure platform, there are 5 sensors, one is obtained 5*22=110 time-frequency characteristics.Table
4 be the characteristic statistics table for independently building failure platform data after rough set theory feature reduction.
Table 4
The time-frequency characteristics of obtained two parts bearing fault data are divided into training sample set and test sample collection.For beauty
State's Case Western Reserve University data choose 80 groups of data as training sample set, randomly select 10 groups of data wherein as training
Sample set A carries out parameter pre-optimized to more disaggregated models, and remaining 70 groups of data are joined as the more disaggregated models of training sample set B
Number suboptimization again, in addition chooses 20 groups of data as test sample collection.Because having chosen 6 kinds of malfunctions of rolling bearing, sieve
32 characteristic parameters relevant to rolling bearing fault state are selected, the diagnosis for rolling bearing fault state provides reference.Institute
With the input layer number of more disaggregated models based on support vector machines network for 32, output layer number of nodes is 6.
For the experiment porch data independently built, 180 groups of data are chosen as training sample set, randomly select 18 groups of numbers
Parameter pre-optimized is carried out to more disaggregated models according to as training sample set A, remaining 162 groups of data are as more points of training sample set B
Class model carries out parameter suboptimization again, in addition selects 20 groups of data as test set.Due to obtaining 5 kinds of malfunctions, screen
69 characteristic parameters relevant to rolling bearing fault state, the diagnosis for rolling bearing fault state provide reference out.So
The input layer number of more disaggregated models based on support vector machines network is 69, and output layer number of nodes is 5.Wherein, training sample
This collection is used for the training of more disaggregated models, and test sample collection is for testing more disaggregated models, with statistical classification accuracy rate.
Table 5 is the parameter setting in the present embodiment in Retrieval method.
| μmax | μmin | c1 | c2 |
| 0.9 | 0.1 | 40 | 10 |
Table 5
Technical effect in order to better illustrate the present invention, using traditional genetic algorithm optimization based on support vector machines
More disaggregated models of network as a comparison method and it is of the invention based on Retrieval method optimization based on support vector machines network
More disaggregated models carry out many experiments together, count fault diagnosis accuracy rate and fault diagnosis operation duration.Table 6 is this reality
Apply in example the present invention and control methods respectively for U.S.'s Case Western Reserve University data and for autonomic nerve way platform data therefore
Hinder diagnostic result statistical form.
Table 6
Fig. 3 is that the present invention and control methods are accurate for the fault diagnosis of U.S.'s Case Western Reserve University data in the present embodiment
Rate statistical chart.Fig. 4 is that the present invention and control methods consume the fault diagnosis of U.S.'s Case Western Reserve University data in the present embodiment
When comparison diagram.Fig. 5 be in the present embodiment the present invention and control methods for autonomic nerve way platform data fault diagnosis accuracy rate
Statistical chart.Fig. 6 is that the present invention and control methods compare the fault diagnosis time-consuming of autonomic nerve way platform data in the present embodiment
Figure.
From table 6, Fig. 3 into Fig. 6 as can be seen that the present invention shows higher accuracy rate, together for two kinds of fault datas
When in many experiments accuracy rate have certain stability.Compared with U.S.'s Case Western Reserve University data, autonomic nerve way platform
Sample size it is bigger, under conditions of big data large sample, the present invention using Retrieval method optimization be based on support vector machines
More disaggregated models of network are with faster diagnosis performance and have certain generalization ability, thus, it is indicated, that of the invention
The more disaggregated models proposed can obtain higher diagnosis within the shorter time, practical in rotary machinery fault diagnosis field
Effectively.
Although the illustrative specific embodiment of the present invention is described above, in order to the technology of the art
Personnel understand the present invention, it should be apparent that the present invention is not limited to the range of specific embodiment, to the common skill of the art
For art personnel, if various change the attached claims limit and determine the spirit and scope of the present invention in, these
Variation is it will be apparent that all utilize the innovation and creation of present inventive concept in the column of protection.
Claims (3)
1. a kind of rotary machinery fault diagnosis method based on Retrieval method Support Vector Machines Optimized, which is characterized in that including
Following steps:
S1: several length are intercepted at random from the working signal of different sensors under the R kind malfunction of rotating machinery respectively
For the working signal L of Ms,n(m), wherein s=1,2 ..., S, S indicate number of sensors, n=1,2 ..., Ns, NsIt indicates to come from s
The working signal quantity of a sensor, m=0,1 ..., M-1, and M > T, T indicate the period of working signal；Remember every section of work
Signal Ls,n(m) corresponding label is Ys,n, label Ys,nFor the malfunction of the corresponding rotating machinery of identification work signal, Ys,n
=1 ..., R；
For every kind of malfunction every section of working signal carry out time frequency analysis processing with extract its temporal signatures, frequency domain character and
Time and frequency domain characteristics, the quantity of every kind of feature determine according to actual needs, using these features as initial characteristics, then for difference
The initial characteristics of the working signal of sensor carry out reduction, and the feature quantity that s-th of sensor initial characteristics reduction of note obtains is
Ds, it is finally total to obtain K feature, whereinRemember each working signal Ls,n(m) corresponding feature vector, Xs,n
=(Xs,n(1),Xs,n(2),…,Xs,n(Ds)), wherein Xs,n(d) working signal L is indicateds,n(m) value of d-th of feature, d=in
1,2,…,Ds；The working signal feature vector of each sensor is classified according to malfunction label, obtains s-th of biography
Feature vector set φ of the working signal of sensor under r kind malfunctions,r, r=1,2 ..., R.
S2: for each malfunction, the training sample of the malfunction is obtained in accordance with the following methods respectively: for the event of r kind
Barrier state, from corresponding S feature vector set φs,rIt is middle randomly select respectively a feature vector, X 's,r, then according to sensing
Device serial number combines to obtain an assemblage characteristic vector Z=(X '1,r,X′2,r,…,X′S,r)=(z1,z2,…,zK), zkIndicate combination
K-th of element of feature vector Z, k=1,2 ..., K, the corresponding malfunction label of assemblage characteristic vector Z are r, that is, constitute one
A training sample；Above procedure is repeated, several training samples are obtained respectively to every kind of malfunction；
All training samples are divided into two parts, a part is used as training sample set A, and another part is as training sample set B, often
The training sample that a training sample is concentrated includes all malfunctions；
S3: more disaggregated models of the building based on support vector machines network, input are assemblage characteristic vector, are exported as malfunction
Label；
S4: using training sample set A as training sample, using Retrieval method to more classification moulds based on support vector machines network
The kernel function and penalty factor of type carry out the processing of parameter pre-optimized；
S5: the kernel function and penalty factor that step S4 parameter pre-optimized is obtained are as more classification based on support vector machines network
The kernel function of model and the initial value of penalty factor, using training sample set B as training sample, using Retrieval method to being based on
The kernel function and penalty factor of more disaggregated models of support vector machines network carry out parameter optimization, obtain final more classification
Model；
S6: when rotating machinery breaks down, S sensor is used to collect S length as the working signal of MFrom
In extract K feature, form assemblage characteristic vectorIt is input to trained more points of step S5
In class model, diagnostic result is obtained.
2. rotary machinery fault diagnosis method according to claim 1, which is characterized in that training sample in the step S2
It divides method particularly includes:
For each malfunction, the average vector of its all training sample is calculated, each training sample is then calculated and this is flat
Training sample is ranked up by the distance of equal vector from big to small according to distance, and training sample is added in Q training sample before selection
Collect A, wherein determine according to actual needs, training sample set A is added in remaining training sample to the value of Q.
3. rotary machinery fault diagnosis method according to claim 1, which is characterized in that described to use Retrieval method pair
The kernel function kernel function and penalty factor of more disaggregated models based on support vector machines network carry out the specific steps of parameter optimization
Include:
S4.1: kernel function and penalty factor to more disaggregated models based on support vector machines network carry out real coding；
S4.2: using the real coding of kernel function and penalty factor as the individual in Retrieval method population, initialization population, kind
Group's size is denoted as Q；
S4.3: each individual in population, using its corresponding kernel function and penalty factor to based on support vector machines network
More disaggregated models be configured, cross validations, the nicety of grading that will be obtained are carried out to the more disaggregated models using training sample
As individual adaptation degree；
S4.4: judge whether otherwise meeting termination iterated conditional enters step S4.6 if it is satisfied, then entering step S4.5；
S4.5: selecting optimum individual from current population, using its corresponding kernel function and penalty factor as parameter optimization result；
S4.6: the individual in population is selected；
S4.7: the degree of certainty μ of each parent individuality selected in step S4.6 is calculated by linear function；Then cloud is calculated
The numerical characteristic of model, including cloud it is expected Ex, entropy EnWith super entropy He；Further according to three of the degree of certainty and cloud model that are calculated
Numerical characteristic executes Y condition cloud generator and obtains a pair of of offspring, calculates and each of selects in obtained offspring and step S4.6
The individual adaptation degree of parent individuality selects preferably Q individual；
S4.8: mutation operation, return step S4.3 are carried out to the Q individual that step S4.7 is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910318220.1A CN110082136B (en) | 2019-04-19 | 2019-04-19 | Rotary machine fault diagnosis method based on cloud genetic algorithm optimization support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910318220.1A CN110082136B (en) | 2019-04-19 | 2019-04-19 | Rotary machine fault diagnosis method based on cloud genetic algorithm optimization support vector machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110082136A true CN110082136A (en) | 2019-08-02 |
| CN110082136B CN110082136B (en) | 2021-03-30 |
Family
ID=67415570
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910318220.1A Active CN110082136B (en) | 2019-04-19 | 2019-04-19 | Rotary machine fault diagnosis method based on cloud genetic algorithm optimization support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110082136B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111340238A (en) * | 2020-03-12 | 2020-06-26 | 中南大学 | Fault diagnosis method, device, equipment and storage medium of industrial system |
| CN111460701A (en) * | 2020-03-09 | 2020-07-28 | 中海油田服务股份有限公司 | Fault diagnosis model training method and device |
| CN112507790A (en) * | 2020-11-03 | 2021-03-16 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on difference evolution |
| CN112598036A (en) * | 2020-12-14 | 2021-04-02 | 湖北工业大学 | Network traffic classification method and system |
| CN112668446A (en) * | 2020-12-24 | 2021-04-16 | 常州大学 | Flower pollination algorithm-based method for monitoring wear state of micro milling cutter by optimizing SVM (support vector machine) |
| CN113139251A (en) * | 2021-04-23 | 2021-07-20 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| WO2021243724A1 (en) * | 2020-06-05 | 2021-12-09 | 北京嘀嘀无限科技发展有限公司 | Image processing method, electronic device, vehicle traveling data recorder, and server |
| CN117970091A (en) * | 2023-12-29 | 2024-05-03 | 广东电网有限责任公司江门供电局 | Fault monitoring method based on multi-classification algorithm |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464964A (en) * | 2007-12-18 | 2009-06-24 | 同济大学 | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis |
| CN105444923A (en) * | 2015-11-18 | 2016-03-30 | 浙江工业大学 | Mechanical temperature instrument error prediction method based on genetic-algorithm optimized least square support vector machine |
| CN106096771A (en) * | 2016-06-07 | 2016-11-09 | 广州供电局有限公司 | The parameter selection method of support vector machine and system and load forecasting method and system |
| CN106777606A (en) * | 2016-12-02 | 2017-05-31 | 上海电机学院 | A kind of gearbox of wind turbine failure predication diagnosis algorithm |
| CN106920006A (en) * | 2017-02-23 | 2017-07-04 | 北京工业大学 | A kind of subway station air conditioning energy consumption Forecasting Methodology based on ISOA LSSVM |
| CN108871762A (en) * | 2018-06-29 | 2018-11-23 | 广东工业大学 | A kind of gearbox of wind turbine method for diagnosing faults |
| CN109583520A (en) * | 2018-12-27 | 2019-04-05 | 云南电网有限责任公司玉溪供电局 | A kind of state evaluating method of cloud model and genetic algorithm optimization support vector machines |
-
2019
- 2019-04-19 CN CN201910318220.1A patent/CN110082136B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464964A (en) * | 2007-12-18 | 2009-06-24 | 同济大学 | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis |
| CN105444923A (en) * | 2015-11-18 | 2016-03-30 | 浙江工业大学 | Mechanical temperature instrument error prediction method based on genetic-algorithm optimized least square support vector machine |
| CN106096771A (en) * | 2016-06-07 | 2016-11-09 | 广州供电局有限公司 | The parameter selection method of support vector machine and system and load forecasting method and system |
| CN106777606A (en) * | 2016-12-02 | 2017-05-31 | 上海电机学院 | A kind of gearbox of wind turbine failure predication diagnosis algorithm |
| CN106920006A (en) * | 2017-02-23 | 2017-07-04 | 北京工业大学 | A kind of subway station air conditioning energy consumption Forecasting Methodology based on ISOA LSSVM |
| CN108871762A (en) * | 2018-06-29 | 2018-11-23 | 广东工业大学 | A kind of gearbox of wind turbine method for diagnosing faults |
| CN109583520A (en) * | 2018-12-27 | 2019-04-05 | 云南电网有限责任公司玉溪供电局 | A kind of state evaluating method of cloud model and genetic algorithm optimization support vector machines |
Non-Patent Citations (4)
| Title |
|---|
| 冯春生: "基于多源不确定信息融合的数控机床滚动轴承故障诊断方法与实验研究", 《中国优秀硕士学位论文全文数据库工程科技I辑》 * |
| 潘美芹 等: "基于实数编码遗传算法的改进支持向量机", 《计算机研究与发展》 * |
| 濮永仙: "实数编码遗传算法与支持向量机在烟草病害识别中的应用", 《江苏农业科学》 * |
| 赵卓 等: "基于改进云遗传算法的多跳自组网故障诊断", 《科技通报》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111460701A (en) * | 2020-03-09 | 2020-07-28 | 中海油田服务股份有限公司 | Fault diagnosis model training method and device |
| CN111340238A (en) * | 2020-03-12 | 2020-06-26 | 中南大学 | Fault diagnosis method, device, equipment and storage medium of industrial system |
| CN111340238B (en) * | 2020-03-12 | 2023-12-15 | 中南大学 | Fault diagnosis method, device, equipment and storage medium of industrial system |
| WO2021243724A1 (en) * | 2020-06-05 | 2021-12-09 | 北京嘀嘀无限科技发展有限公司 | Image processing method, electronic device, vehicle traveling data recorder, and server |
| CN112507790A (en) * | 2020-11-03 | 2021-03-16 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on difference evolution |
| CN112507790B (en) * | 2020-11-03 | 2023-09-29 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on differential evolution |
| CN112598036A (en) * | 2020-12-14 | 2021-04-02 | 湖北工业大学 | Network traffic classification method and system |
| CN112668446A (en) * | 2020-12-24 | 2021-04-16 | 常州大学 | Flower pollination algorithm-based method for monitoring wear state of micro milling cutter by optimizing SVM (support vector machine) |
| CN113139251A (en) * | 2021-04-23 | 2021-07-20 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| CN113139251B (en) * | 2021-04-23 | 2023-03-14 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| CN117970091A (en) * | 2023-12-29 | 2024-05-03 | 广东电网有限责任公司江门供电局 | Fault monitoring method based on multi-classification algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110082136B (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110082136A (en) | Rotary machinery fault diagnosis method based on Retrieval method Support Vector Machines Optimized | |
| CN110334391B (en) | Automatic planning method for collecting circuit of multi-dimensional constraint wind power plant | |
| CN104751229B (en) | Bearing fault diagnosis method capable of recovering missing data of back propagation neural network estimation values | |
| CN105184084B (en) | Method and system for predicting fault type of electric power metering automation terminal | |
| CN108171263A (en) | Based on the Fault Diagnosis of Roller Bearings for improving variation mode decomposition and extreme learning machine | |
| CN109102032A (en) | A kind of pumping plant unit diagnostic method based on depth forest and oneself coding | |
| CN107991097A (en) | A kind of Method for Bearing Fault Diagnosis based on multiple dimensioned symbolic dynamics entropy | |
| CN106447039A (en) | Non-supervision feature extraction method based on self-coding neural network | |
| CN107101828A (en) | The bearing residual life Forecasting Methodology that a kind of encoder health degree is evaluated | |
| CN102768701B (en) | High-voltage switch cabinet insulator electric field optimization method based on quantum genetic algorithm | |
| CN113177357B (en) | Transient stability assessment method for power system | |
| CN106021771A (en) | Method and device for diagnosing faults | |
| CN106650982A (en) | Depth learning power prediction method based on multi-point NWP | |
| CN107133632A (en) | A kind of wind power equipment fault diagnosis method and system | |
| CN105843733B (en) | A kind of method for testing performance and device of big data platform | |
| Swarup et al. | ANN approach assesses system security | |
| CN106295857A (en) | A kind of ultrashort-term wind power prediction method | |
| CN103268222A (en) | Method and system for arranging digital human-machine interface monitoring units | |
| Zheng et al. | Real-time transient stability assessment based on deep recurrent neural network | |
| CN106768933A (en) | A kind of wind power generation unit blade method for diagnosing faults based on genetic algorithm | |
| CN113887729A (en) | Fault diagnosis method for low-voltage power line carrier communication system | |
| Saragih et al. | Application of extreme learning machine and modified simulated annealing for jatropha curcas disease identification | |
| CN106526465A (en) | High-voltage circuit breaker fault intelligent diagnosis method based on improved fuzzy Petri network | |
| CN111062569A (en) | Low-current fault discrimination method based on BP neural network | |
| CN109886316B (en) | Transformer state parameter combination prediction method based on cloud system similarity weight distribution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |